Anomaly Detection in Network Traffic for Proactive Security Threat Identification Using Improved Gated Recurrent Unit

Anomaly detection in network traffic determines the unusual patterns which diverge from normal behavior by representing potential system faults and security threats. It assists in identifying cyberattacks, performance issue, and network intrusions through identifying traffic flow and deviations. Continuous monitoring makes rapid response and effective threat detection which reduces service disruption and damage. Existing network security faces challenged in detecting the anomalous traffic patterns which is caused by evolving cyberthreats leading to data breaches. Therefore, effective anomaly detection is essential to improve network security, ensure network stability, and prevent intrusion. Hence, this research proposes improved Gated Recurrent Unit (GRU) with feature extraction of sequential patterns and adaptive mechanism to detect the anomaly in network traffic effectively. Initially, the gathered data is normalized using min-max normalization to standardize the data and then one-hot encoding is applied for encoding the labels effectively. Then, the improved GRU is used to detect the anomaly in network traffic which enhance high security. The proposed improved GRU obtains a high accuracy compared to existing methods like U-Net based and Temporal convolutional network (TCN) using KDD99 and the CSE-CIC-IDS2018 datasets.